Skip to content

Latest commit

 

History

History
143 lines (95 loc) · 5.73 KB

documentation.md

File metadata and controls

143 lines (95 loc) · 5.73 KB
Raw

Mozilla Traffic Cop Documentation

Traffic Cop places visitors into A/B/x cohorts, and either performs a redirect or executes a developer-specified JavaScript function.

How it works

After verifying the supplied configuration, Traffic Cop then generates a random number to choose a variation based on the supplied cohort percentages. If the supplied variations do not target 100% of visitors, the no-variation value may be chosen.

Traffic Cop does not use cookies to record or store data of any kind. It also does not store, send or transmit any kind of experiment data for analysis. It simply performs the task of displaying different experiment variations upon page load. It is up to you to record experiment data using whatever your standard website analytics tools may be.

Note that because Traffic Cop does not use cookies, this does mean that repeat visits to the canonical URL might see a different variation to the one they saw previously during the course of an experiment running.

Type A: Callback

If the instance of Traffic Cop is provided a customCallback in the config, the function supplied will be passed the variation value and executed.

Any further functionality is handled by the customCallback.

For performance, and to not require any kind of document.ready code, it is recommended that "Type A" instances of Traffic Cop have their .js files placed at the bottom of the page, just before the closing </body> tag.

// example configuration for a callback function experiment
function myCallback(variation) {
    console.log('The chosen variation was ' + variation);
    // and then change button color based on variation chosen...
}

var lou = new TrafficCop({
  customCallback: myCallback,
  variations: {
    ‘a’: 10.25,
    ‘b’: 20.2,
    'c': 0.1,
    'd': 0.55
  }
});

lou.init();

Type B: Redirect

If the instance of Traffic Cop is not provided a customCallback, and if the chosen variation is not no-variation, the visitor is redirected to the current URL with the chosen variation appended to the query string.

Any further functionality is handled by the application. (Usually loading different HTML/JS/CSS based on the query parameter.)

If the chosen variation is no-variation, no redirect will occur - i.e. the user is not chosen to participate in the experiment.

Any query string parameters present when a user initially lands on a page will be propagated to the variation redirect.

To avoid content flicker, it is recommended that "Type B" instances of Traffic Cop have their .js files placed in the <head> of the page.

// example configuration for a redirect experiment
var wiggum = new TrafficCop({
  variations: {
    ‘v=1’: 0.15,
    ‘v=2’: 30,
    'v=3': 10.6
  }
});

wiggum.init();

Note that the variations for redirect experiments contain a key=value pair. This is not required, but does result in nicer redirect URLs, e.g. https://www.example.com/?v=2 instead of https://www.example.com/?2.

Advanced: Type B + Type A

It is possible to have both types of experiments running on the same page for the same audience. The general setup would be:

  1. Place the redirect experiment in the <head> of the page. This will take precedence.
  2. Place the callback experiment at the end of the page. This will execute regardless of query string values.

Check out the demo for a live example of this setup.

How a variation is chosen

Variations are sorted in the order provided, and percentages are tallied to create tiers. Take the following config:

var rex = new TrafficCop({
    variations: {
        'v=a': 15,
        'v=b': 0.25,
        'v=c': 25.6
    }
});

The implied tiers would be:

  1. v=a: 1-15
  2. v=b: 15.01 - 15.25
  3. v=c: 15.26 - 40.31
  4. (no variation chosen): 40.32-100

If the random number generated was 12.6, the user would be redirected to ?v=a. A value of 15.2 would send the user ?v=b, 15.6 to ?v=c, and anything above 40.31 would result in no redirect.

Note that Traffic Cop supports percentages into the hundredths, but no smaller.

Configuration

Each instance of a Traffic Cop requires one main configuration property:

  • A variations object that lists all variations along with the associated percent chance of being chosen

An implementation for a redirect experiment might look like:

var eddie = new TrafficCop({
    variations: {
        'v=1': 12.2,
        'v=2': 0.13,
        'v=3': 11.45
    }
});

eddie.init();

In the above example, the test will have 3 variations and will target a total of 23.78% of visitors. There will also be a 76.22% chance that no-variation is chosen.

Implementation

Traffic Cop requires two JavaScript files:

  1. @mozmeao/trafficcop (install via NPM)
  2. A custom .js file to configure and initialize an instance of Traffic Cop (and perhaps contain a callback function).

You can import the NPM package directly into your custom .js file to bundle everything together, using require or import:

  • import TrafficCop from '@mozmeao/trafficcop';
  • const TrafficCop = require('@mozmeao/trafficcop');

Considerations

  1. To prevent search engines from indexing a variation URL, we recommend adding a <link rel="canonical"> to the <head> of your experiment pages that points to the URL without any variation parameters. For example, all variations for www.example.com/product should have the following tag:

    <link rel="canonical" href="http://www.example.com/product">

  2. Concatenate and minify experiment-specific files before sending to production. This will reduce your file size by about 70%!

  3. Make sure your visitors consent to analytics before recording experiment data.